controller in question generates frames with bad IP checksum value
if packets contain IP options. For instance, packets generated by
ping(8) with record route option have wrong IP checksum value. The
controller correctly computes checksum for normal TCP/UDP packets
though.
There are two known RTL8168/8111C variants in market and the issue
I observed happened on RL_HWREV_8168C_SPIN2. I'm not sure
RL_HWREV_8168C also has the same issue but it would be better to
assume it has the same issue since they shall share same core.
RTL8102E which is supposed to be released at the time of
RTL8168/8111C announcement does not have the issue.
Tested by: Konstantin V. Krotov ( kkv <> insysnet dot ru )
This isn't strictly required to TX (at least non-agg and non-HT40,
non-short-GI) frames; but as it needs to be done anyway, just get
it done.
Linux ath9k uses the rate scenario style path for -all- packets,
legacy or otherwise. This code does much the same.
Beacon TX still uses the legacy, non-rate-scenario TX descriptor
setup. Ath9k also does this.
This 11n rate scenario path is only called for chips in the AR5416
HAL; legacy chips use the previous interface for TX'ing.
A-MPDU RX interferes with packet retransmission/reordering.
In local testing, I was seeing A-MPDU being negotiated and then
not used by the AP sending frames to the STA; the STA would then
treat non A-MPDU frames that are retransmits as out of the window
and get plain confused.
The hardware RX status descriptor has a "I'm part of an aggregate"
bit; so this should eventually be tested and then punted to the
A-MPDU reorder handling only if it has this bit set.
get properties from the parent. The parent is in fact the FDT bus
itself and will therefore not have the properties we're looking
for.
Sponsored by: Juniper Networks
1) They don't use the giant "MAX_CPU" define and instead
are allocated dynamically based on mp_ncpus
2) Will zero with the netstat -z -s -p sctp
3) Will be properly handled by both the sctp_init and finish
(the multi-net stuff was incorrectly bzero'ing in sctp_init
the wrong size.. the bzero is now moved to the right places).
And of course the free is put in at the very end.
MFC after: 3 Months
threads. These serve as input threads and are queued
packets based on the V-tag number. This is similar to
what a modern card can do with queue's for TCP... but
alas modern cards know nothing about SCTP.
MFC after: 3 months (maybe)
make use of the aac_ioctl_event callback, if aac_alloc_command fails. This
can end up in an infinite loop in the while loop in aac_release_command.
Further investigation into the issue mentioned by Scott Long [1] will be
necessary.
[1] http://lists.freebsd.org/pipermail/freebsd-current/2007-October/078740.html
should_yield(). Use this in various places. Encapsulate the common
case of check-and-yield into a new function maybe_yield().
Change several checks for a magic number of iterations to use
should_yield() instead.
MFC after: 1 week
never committed from p4 dtrace branch.
[The correct include path is referenced from every opensolaris compat
consumer's module Makefile, so it doesn't serve any purpose anyway.]
Reported by: arundel on freebsd-hackers@ via clang
Approved by: kib (mentor)
MFC after: 1 week
2) Add separate max-bursts for retransmit and hb. These
are set to sysctlable values but not settable via the
socket api. This makes sure we don't blast out HB's or
fast-retransmits.
3) Determine on the first data transmission on a net if
its local-lan (by being under or over a RTT). This
can later be used to think about different algorithms
based on locallan vs big-i (experimental)
4) The cwnd should NOT be allowed to grow when an ECNEcho
is seen (TCP has this same bug). We fix this in SCTP
so an ECNe being seen prevents an advance of cwnd.
5) CWR's should not be sent multiple times to the
same network, instead just updating the TSN being
transmitted if needed.
MFC after: 1 Month
so that future allocations start with most recently allocated block
rather than the beginning of the filesystem.
- Fix ext2_alloccg() to properly scan for 8 block chunks that are not
aligned on 8-bit boundaries. Previously this was causing new blocks
to be allocated in a highly fragmented fashion (block 0 of a file at
lbn N, block 1 at lbn N + 8, block 2 at lbn N + 16, etc.).
- Cosmetic tweaks to the currently-disabled fancy realloc sysctls.
PR: kern/153584
Discussed with: bde
Tested by: Pedro F. Giffuni giffunip at yahoo, Zheng Liu (lz)
The clock_t type in OpenSolaris is long (int64_t on amd64).
On FreeBSD clock_t is int32_t. The clock_t type is used in several places
in the ZFS code to store system uptime in milliseconds ("seconds * hz").
With hz=1000 we have a 32-bit integer overflow in 24 days, 20 hours,
31 minutes and 23.648 seconds. This has a user reported negative impact
on l2arc_feed_thread() and may cause unexpected results from other functions
using clock_t.
Reported by: Artem Belevich <fbsdlist@src.cx> on freebsd-fs@
MFC after: 1 week
collect phases. The unp_discard() function executes
unp_externalize_fp(), which might make the socket eligible for gc-ing,
and then, later, taskqueue will close the socket. Since unp_gc()
dropped the list lock to do the malloc, close might happen after the
mark step but before the collection step, causing collection to not
find the socket and miss one array element.
I believe that the race was there before r216158, but the stated
revision made the window much wider by postponing the close to
taskqueue sometimes.
Only process as much array elements as we find the sockets during
second phase of gc [1]. Take linkage lock and recheck the eligibility
of the socket for gc, as well as call fhold() under the linkage lock.
Reported and tested by: jmallett
Submitted by: jmallett [1]
Reviewed by: rwatson, jeff (possibly)
MFC after: 1 week
top 8 bits of the 32 bit signal bit field space for internal use. These private
signals should not be leaked outside of a module.
Given that many algorithm modules use the NewReno hook functions to simplify
their implementation, the obvious place such a leak would show up is in the
NewReno cong_signal hook function.
- Show the full number of significant bits in the signal type definitions in
<netinet/cc.h>.
- Add a bitmask to simplify figuring out if a given signal is in the private or
public bit range.
- Add a sanity check in newreno_cong_signal() to ensure private signals are not
being leaked into the hook function.
Sponsored by: FreeBSD Foundation
Discussed with: David Hayes <dahayes at swin edu au>
MFC after: 1 week
X-MFC with: r215166
The AR5416 and later TX descriptors have new fields for supporting
11n bits (eg 20/40mhz mode, short/long GI) and enabling/disabling
RTS/CTS protection per rate.
These functions will be responsible for initialising the TX descriptors
for the AR5416 and later chips for both HT and legacy frames.
Beacon frames will remain using the non-11n TX descriptor setup for now;
Linux ath9k does much the same.
Note that these functions aren't yet used anywhere; a few more framework
changes are needed before all of the right rate information is available
for TX.
algorithm described in the paper "Improved coexistence and loss tolerance for
delay based TCP congestion control" by Hayes and Armitage. It is implemented as
a kernel module compatible with the recently committed modular congestion
control framework.
CHD enhances the approach taken by the Hamilton-Delay (HD) algorithm to provide
tolerance to non-congestion related packet loss and improvements to coexistence
with loss-based congestion control algorithms. A key idea in improving
coexistence with loss-based congestion control algorithms is the use of a shadow
window, which attempts to track how NewReno's congestion window (cwnd) would
evolve. At the next packet loss congestion event, CHD uses the shadow window to
correct cwnd in a way that reduces the amount of unfairness CHD experiences when
competing with loss-based algorithms.
In collaboration with: David Hayes <dahayes at swin edu au> and
Grenville Armitage <garmitage at swin edu au>
Sponsored by: FreeBSD Foundation
Reviewed by: bz and others along the way
MFC after: 3 months
function; which will be later used by the TX path to determine
whether to use the extended features or not.
* Break out the descriptor chaining logic into a separate function;
again so it can be switched out later on for the 11n version when
needed.
* Refactor out the encryption-swizzling code that's common in the
raw and normal TX path.
algorithm based on the paper "A strategy for fair coexistence of loss and
delay-based congestion control algorithms" by Budzisz, Stanojevic, Shorten and
Baker. It is implemented as a kernel module compatible with the recently
committed modular congestion control framework.
HD uses a probabilistic approach to reacting to delay-based congestion. The
probability of reducing cwnd is zero when the queuing delay is very small,
increasing to a maximum at a set threshold, then back down to zero again when
the queuing delay is high. Normal operation keeps the queuing delay below the
set threshold. However, since loss-based congestion control algorithms push the
queuing delay high when probing for bandwidth, having the probability of
reducing cwnd drop back to zero for high delays allows HD to compete with
loss-based algorithms.
In collaboration with: David Hayes <dahayes at swin edu au> and
Grenville Armitage <garmitage at swin edu au>
Sponsored by: FreeBSD Foundation
Reviewed by: bz and others along the way
MFC after: 3 months
based on the paper "TCP Vegas: end to end congestion avoidance on a global
internet" by Brakmo and Peterson. It is implemented as a kernel module
compatible with the recently committed modular congestion control framework.
VEGAS uses network delay as a congestion indicator and unlike regular loss-based
algorithms, attempts to keep the network operating with stable queuing delays
and no congestion losses. By keeping network buffers used along the path within
a set range, queuing delays are kept low while maintaining high throughput.
In collaboration with: David Hayes <dahayes at swin edu au> and
Grenville Armitage <garmitage at swin edu au>
Sponsored by: FreeBSD Foundation
Reviewed by: bz and others along the way
MFC after: 3 months
The higher levels (net80211, if_ath, ath_rate) need this to make correct
choices about what MCS capabilities to advertise and what MCS rates are
able to be TXed.
In summary:
* AR5416 - 2/3 antennas, 2x2 streams
* AR9160 - 2/3 antennas, 2x2 streams
* AR9220 - 2 antennas, 2x2 sstraems
* AR9280 - 2 antennas, 2x2 streams
* AR9285 - 2 antennas but with antenna diversity, 1x1 stream
- SMBus Controller
- SATA Controller
- HD Audio Controller
- Watchdog Controller
Thanks to Seth Heasley (seth.heasley@intel.com) for providing us code.
MFC after 3 days
apply AR8152 v1.0 specific initialization code. Fix this bug by
explicitly reading PCI device revision id via PCI accessor.
Reported by: Gabriel Linder ( linder.gabriel <> gmail dot com )
After inspecting the ath9k source, it seems the AR5416 and later MACs
don't take an explicit RTS/CTS duration. A per-scenario (ie, what multi-
rate retry became) rts/cts control flag and packet duration is provided;
the hardware then apparently fills in whatever details are required.
The per-rate sp/lpack duration calculation just isn't used anywhere
in the ath9k TX packet length calculations.
The burst duration register controls something different; it seems to
be involved with RTS/CTS protection of 11n aggregate frames and is set
via a call to ar5416Set11nBurstDuration().
I've done some light testing with rts/cts protected frames and nothing
seems to break; but this may break said RTS/CTS and CTS-to-self protection.
1) We now remove ECN-Nonce since it will no longer continue as a I-D
2) Eliminate last_tsn_echo, this tied us to an assoc not the net
and thus we were not doing m-homing on the ECN-Echo senders side right.
3) Increment the count going out even if the TSN in lower in the pending
ECN-Echo, this way the receiver knows exactly how many packets were
marked even with network re-ordering
4) Fix so we DO NOT stop doing delayed sack if a ECN Echo is in queue
MFC after: 1 month
try to export as much information as we can match.
Requested on: Debian GNU/kFreeBSD list (debian-bsd lists.debian.org) 2010-12
Tested by: Mats Erik Andersson (mats.andersson gisladisker.se)
MFC after: 10 days
if a scan is running, report if a scan has been started. The return value
itself is not (yet) used anywhere in the tree and it is also not exported
to userspace.
MFC after: 1 month
that represents the host controller. This makes the FDT PCI support
working an a bare-bones manner. This needs a lot more work, of which
the beginning are at the end of the file, compiled-out with #if 0.
The intend being that both the Marvell PCIE and Freescale PCI/PCIX/PCIE
duplicate the same platform-independent domain initialization, that
should be moved into an unified implementation in the FDT code. Handling
of resources requires help from the platform. A unified implementation
allows us to properly support PCI devices listed in the device tree and
configured according to the device tree specification.
Sponsored by: Juniper Networks
already supported nested PICs, but was limited to having a nested
AT-PIC only. With G5 support the need for nested OpenPIC controllers
needed to be added. This was done the wrong way and broke the MPC8555
eval system in the process.
OFW, as well as FDT, describe the interrupt routing in terms of a
controller and an interrupt pin on it. This needs to be mapped to a
flat and global resource: the IRQ. The IRQ is the same as the PCI
intline and as such needs to be representable in 8 bits. Secondly,
ISA support pretty much dictates that IRQ 0-15 should be reserved
for ISA interrupts, because of the internal workins of south bridges.
Both were broken.
This change reverts revision 209298 for a big part and re-implements
it simpler. In particular:
o The id() method of the PIC I/F is removed again. It's not needed.
o The openpic_attach() function has been changed to take the OFW
or FDT phandle of the controller as a second argument. All bus
attachments that previously used openpic_attach() as the attach
method of the device I/F now implement as bus-specific method
and pass the phandle_t to the renamed openpic_attach().
o Change powerpc_register_pic() to take a few more arguments. In
particular:
- Pass the number of IPIs specificly. The number of IRQs carved
out for a PIC is the sum of the number of int. pins and IPIs.
- Pass a flag indicating whether the PIC is an AT-PIC or not.
This tells the interrupt framework whether to assign IRQ 0-15
or some other range.
o Until we implement proper multi-pass bus enumeration, we have to
handle the case where we need to map from PIC+pin to IRQ *before*
the PIC gets registered. This is done in a similar way as before,
but rather than carving out 256 IRQs per PIC, we carve out 128
IRQs (124 pins + 4 IPIs). This is supposed to handle the G5 case,
but should really be fixed properly using multiple passes.
o Have the interrupt framework set root_pic in most cases and not
put that burden in PIC drivers (for the most part).
o Remove powerpc_ign_lookup() and replace it with powerpc_get_irq().
Remove IGN_SHIFT, INTR_INTLINE and INTR_IGN.
Related to the above, fix the Freescale PCI controller driver, broken
by the FDT code. Besides not attaching properly, bus numbers were
assigned improperly and enumeration was broken in general. This
prevented the AT PIC from being discovered and interrupt routing to
work properly. Consequently, the ata(4) controller stopped functioning.
Fix the driver, and FDT PCI support, enough to get the MPC8555CDS
going again. The FDT PCI code needs a whole lot more work.
No breakages are expected, but lackiong G5 hardware, it's possible
that there are unpleasant side-effects. At least MPC85xx support is
back to where it was 7 months ago -- it's amazing how badly support
can be broken in just 7 months...
Sponsored by: Juniper Networks
1) ECN was on an association basis, this is incorrect and
will not work with CMT or for that matter if the user
is sending to multiple addresses. This commit makes
ECN on a per path basis.
2) Adopt the new format for the ECN internet draft. This also
maintains compatability with old format chunks as well.
3) Keep track of the real time of a RTT down to micro seconds.
For some future conditional features (for like a data center
this is good information to have).
MFC after: 1 month
MAP_STACK_* entries. (See r71983 and r74235.)
In some cases, performing this call to vm_map_simplify_entry() halves the
number of vm map entries used by the Sun JDK.
Each different radio chipset has a different "good" range of CCA
(clear channel access) parameters where, if you write something
out of range, it's possible the radio will go deaf.
Also, since apparently occasionally reading the NF calibration
returns "wrong" values, so enforce those limits on what is being
written into the CCA register.
Write a default value if there's no history available.
This isn't the case right now but it may be later on when "off-channel"
scanning occurs without init'ing or changing the NF history buffer.
(As each channel may have a different noise floor; so scanning or
other off-channel activity shouldn't affect the NF history of
the current channel.)
* I messed up a couple of things in if_athvar.h; so fix that.
* Undo some guesswork done in ar5416Set11nRateScenario() and introduce a
flags parameter which lets the caller set a few things. To begin with,
this includes whether to do RTS or CTS protection.
* If both RTS and CTS is set, only do RTS. Both RTS and CTS shouldn't be
set on a frame.
There's two reasons for this:
* the raw and non-raw TX path shares a lot of duplicate code which should be
refactored;
* the 11n-ready chip TX path needs a little reworking.
receive processing.
Remove unnecessary restrictions on the mbuf chain length built during an
LRO receive. This restriction was copied from the Linux netfront driver
where the LRO implementation cannot handle more than 18 discontinuities.
The FreeBSD implementation has no such restriction.
MFC after: 1 week
This will be used for Data Center congestion
control, we won't want to engage it in the
ECN code unless we KNOW that the RTT is less
than 500us.
MFC after: 1 week
sends were being accounted for. The
counting was such that we counted only
when we queued a chunk, not when we sent it.
Now keep an additional counter for queuing and
one for sending.
MFC after: 1 week
be used by linuxolator itself.
Move linux_wait4() to MD path as it requires native struct
rusage translation to struct l_rusage on linux32/amd64.
MFC after: 1 Month.
each of the threads needs more while current pool of the buffers is
exhausted, then neither thread can make progress.
Switch to nowait allocations after we got first buffer already.
Reported by: az
Reviewed by: alc (previous version)
Tested by: pho
MFC after: 1 week
The fdcheckstd() function makes sure fds 0, 1 and 2 are open by opening
/dev/null. If this fails (e.g. missing devfs or wrong permissions),
fdcheckstd() will return failure and the process will exit as if it received
SIGABRT. The KASSERT is only to check that kern_open() returns the expected
fd, given that it succeeded.
Tripping the KASSERT is most likely if fd 0 is open but fd 1 or 2 are not.
MFC after: 2 weeks
EBR schemes: fat32, ebr, linux-data, linux-raid, linux-swap and
linux-lvm. Add bios-boot GUID and alias for the GPT scheme. It used by
GRUB 2 loader. Also do sorting definitions of types in diskmbr.h
and in g_part.c.
PR: bin/120990, kern/147664
MFC after: 2 weeks
This is just the bare minimum needed to teach ath_rate_sample to try
and handle MCS rates. It doesn't at all attempt to find the best
rate by any means - it doesn't know anything about the MCS rate
relations, TX aggregation or any of the much sexier 11n stuff
that's out there.
It's just enough to transmit 11n frames and handle TX completion.
It shouldn't affect legacy (11abg) behaviour.
Obtained from: rpaulo@
This will eventually be used by rate control modules and by the TX
code for calculating packet duration when handling rts/cts protection.
Obtained from: sam@, rpaulo@, linux ath9k
members, thus making a signed extension of 32 bit register
context. If the register is not touched in usermode between
return from signal and next syscall entry, the sign-extension
part of 64bit register is not cleared, causing
linux32_fetch_syscall_args() to read wrong values.
Use unsigned type for the registers in the linux sigcontext.
Reported by: Jacob Frelinger <jacob.frelinger duke edu>, arundel
In collaboration with: dchagin
MFC after: 1 week
covering the whole page, free the page. Otherwise, clear the region and
mark it clean. Not marking the page dirty could reinstantiate cleared
data, but it is allowed by BIO_DELETE specification and saves unneeded
write to swap.
Reviewed by: alc
Tested by: pho
MFC after: 2 weeks
- Provide trivial implementation of sf_buf_alloc(), sf_buf_free(),
sf_buf_kva() and sf_buf_page() using direct map for n64.
- uio_machdep.c - use macros so that the direct map will be used in
case of n64.
Reviewed by: imp (earlier version)
Obtained from: jmallett (user/jmallett/octeon)
The defaults enabled three chains on the AR5416 even if the card has two
chains. This restores that and ensures that only the correct TX/RX
chainmasks are used.
When HT modes are enabled, all TX chains will be correctly enabled.
This should now enable analog chain swapping with 2-chain cards.
I'm not sure if this is needed for just the AR5416 or whether
it also applies to AR9160, AR9280 and AR9287 (later on); I'll have
to get clarification.
This, along with an initval change which will appear in a subsequent commit,
fixes bus panics that I have been seing with the AR9220 on a Routerstation Pro
(AR7161 MIPS board.)
Obtained from: Linux ath9k
PR: kern/154220
sbuf_new_for_sysctl(9). This allows using an sbuf with a SYSCTL_OUT
drain for extremely large amounts of data where the caller knows that
appropriate references are held, and sleeping is not an issue.
Inspired by: rwatson
unfunctional. Wiring the user buffer has only been done explicitly
since r101422.
Mark the kern.disks sysctl as MPSAFE since it is and it seems to have
been mis-using the NOLOCK flag.
Partially break the KPI (but not the KBI) for the sysctl_req 'lock'
field since this member should be private and the "REQ_LOCKED" state
seems meaningless now.
the controller has a kind of embedded controller/memory and vendor
applies a large set of magic code via undocumented PHY registers in
device initialization stage. I guess it's a firmware image for the
embedded controller in RTL8105E since the code is too big compared
to other DSP fixups. However I have no idea what that magic code
does and what's purpose of the embedded controller. Fortunately
driver seems to still work without loading the firmware.
While I'm here change device description of RTL810xE controller.
H/W donated by: Realtek Semiconductor Corp.
exact model name is not clear yet. All previous RTL8201 10/100 PHYs
used 0x8201 in MII_PHYIDR2 which in turn makes model number 0x20
but this PHY used new model number 0x08.
capability. One of reason using interrupt taskqueue in re(4) was
to reduce number of TX/RX interrupts under load because re(4)
controllers have no good TX/RX interrupt moderation mechanism.
Basic TX interrupt moderation is done by hardware for most
controllers but RX interrupt moderation through undocumented
register showed poor RX performance so it was disabled in r215025.
Using taskqueue to handle RX interrupt greatly reduced number of
interrupts but re(4) consumed all available CPU cycles to run the
taskqueue under high TX/RX network load. This can happen even with
RTL810x fast ethernet controller and I believe this is not
acceptable for most systems.
To mitigate the issue, use one-shot timer register to moderate RX
interrupts. The timer register provides programmable one-shot timer
and can be used to suppress interrupt generation. The timer runs at
125MHZ on PCIe controllers so the minimum time allowed for the
timer is 8ns. Data sheet says the register is 32 bits but
experimentation shows only lower 13 bits are valid so maximum time
that can be programmed is 65.528us. This yields theoretical maximum
number of RX interrupts that could be generated per second is about
15260. Combined with TX completion interrupts re(4) shall generate
less than 20k interrupts. This number is still slightly high
compared to other intelligent ethernet controllers but system is
very responsive even under high network load.
Introduce sysctl variable dev.re.%d.int_rx_mod that controls amount
of time to delay RX interrupt processing in units of us. Value 0
completely disables RX interrupt moderation. To provide old
behavior for controllers that have MSI/MSI-X capability, introduce
a new tunable hw.re.intr_filter. If the tunable is set to non-zero
value, driver will use interrupt taskqueue. The default value of
the tunable is 0. This tunable has no effect on controllers that
has no MSI/MSI-X capability or if MSI/MSI-X is explicitly disabled
by administrator.
While I'm here cleanup interrupt setup/teardown since re(4) uses
single MSI/MSI-X message at this moment.
with the latest socket API ID. Especially it can be disabled.
Full compliance needs changing the structure used in the
socket option. Since this breaks the API, it will be a
seperate commit which will not be MFCed to stable/8.
MFC after: 3 months.
ath9k does a few different things here during config - if it's an early
AR5416 with two chains, it enables all three chains for calibration and
then restores the chainmask to the original values after initial
calibration has completed.
The reason behind this commit is to begin breaking out the chainmask
configuration for this specific reason; follow-up commits will add
the chainmask restore in the ar5416Reset() routine.
recent PCIe controllers(RTL8102E or later and RTL8168/8111C or
later) supports either 2 or 4 MSI-X messages. Unfortunately vendor
did not publicly release RSS related information yet. However
switching to MSI-X is one-step forward to support RSS.
RTL8111C generated corrupted frames where TCP option header was
broken. All other sample controllers I have did not show such
problem so it could be RTL8111C specific issue. Because there are
too many variants it's hard to tell how many controllers have such
issue. Just disable TSO by default but have user override it.
before checking the validity of the next buffer pointer. Otherwise, the
buffer might be reclaimed after the check, causing iteration to run into
wrong buffer.
Reported and tested by: pho
MFC after: 1 week
* Re-do the structure size/component math to make sure the struct matches
the expected size
* Just to be clear that we care about bitmask ordering, revert my previous
change and instead define that macro if we're on big-endian.
It turns out that the V4K eeprom definitions (used by the AR9285 and
its derivatives) is wrong. These values are at least causing issues
on my AR2427.
With this fix (and initvals in a subsequent commit), the AR2427 behaves
a lot better.
Note - there's still significant drift between the ath9k v4k eeprom
init code (again, used by AR9285 and derivatives) and what's in this
tree. That needs to be investigated and resolved.
Khelp/Hhook KPIs to hook into the TCP stack and maintain a per-connection, low
noise estimate of the instantaneous RTT. ERTT's implementation is robust even in
the face of delayed acknowledgements and/or TSO being in use for a connection.
A high quality, low noise RTT estimate is a requirement for applications such as
delay-based congestion control, for which we will be importing some algorithm
implementations shortly.
In collaboration with: David Hayes <dahayes at swin edu au> and
Grenville Armitage <garmitage at swin edu au>
Sponsored by: FreeBSD Foundation
Reviewed by: bz and others along the way
MFC after: 3 months
table in if_grow(). The order of the SYSINIT's for ifnet state were swapped
so that the various locks were initialized before being used.
Reviewed by: pluknet, bz
MFC after: 2 weeks
prevent sending data when CTS is de-asserted.
In uart_tty_intr(), call uart_tty_outwakeup() when the CTS signal
changed, knowing that uart_tty_outwakeup() will do the right
thing for flow control. This avoids redundant conditionals.
PR: kern/148644
Submitted by: John Wehle <john@feith.com>
MFC after: 3 days
via AHCI-like memory resource at BAR(5). Use it if BIOS was so kind to
allocate memory for that BAR. This allows hot-plug support and connection
speed reporting.
MFC after: 2 weeks
controllers. Experimentation with RTL8102E, RTL8103E and RTL8105E
showed dramatic decrement of TX completion interrupts under high TX
load(e.g. from 147k interrupts/second to 10k interrupts/second)
With this change, TX interrupt moderation is applied to all
controllers except RTL8139C+.
is on the MacIO ones. It appears to be unreliable on all DBDMA-based
controllers for unknown reasons, which should be figured out eventually.
Tested by: Torfinn Ingolfsen
MFC after: 1 week
The linux ath9k driver and (from what I've been told) the atheros reference
driver does this; it then leaves discarding 11n frames to the 802.11 layer.
Whilst I'm here, merge in a fix from ath9k which maintains a turbo register
setting when enabling the 11n register; and remove an un-needed (duplicate)
flag setting.
write to the buffer causes it to overflow. We therefore can't hold the CC list
rwlock over a call to sbuf_printf() for an sbuf configured with SBUF_AUTOEXTEND.
Switch to a fixed length sbuf which should be of sufficient size except in the
very unlikely event that the sysctl is being processed as one or more new
algorithms are loaded. If that happens, we accept the race and may fail the
sysctl gracefully if there is insufficient room to print the names of all the
algorithms.
This should address a WITNESS warning and the potential panic that would occur
if the sbuf call to malloc did sleep whilst holding the CC list rwlock.
Sponsored by: FreeBSD Foundation
Reported by: Nick Hibma
Reviewed by: bz
MFC after: 3 weeks
X-MFC with: r215166
defined. The kernel linker doesn't deal with symbols of
type NOTYPE and typically gives the wrong symbol ($a) for
local symbols.
Obtained from: Juniper Networks, Inc.
The v1 and v3 interfaces returned the whole EEPROM but the v14/v4k
interfaces just returned the base header. There's extra information
outside of that which would also be nice to get access to.
The rxmonitor hook is called on each received packet. This can get very,
very busy as the tx/rx/chanbusy registers are thus read each time a packet
is received.
Instead, shuffle out the true per-packet processing which is needed and move
the rest of the ANI processing into a periodic event which runs every 100ms
by default.
- The mean RTT is updated at the end of each congestion epoch, but if we switch
to congestion avoidance within the first epoch (e.g. if ssthresh was primed
from the hostcache), we'll trigger a divide by zero panic in
cubic_ack_received(). Set the mean to the min in cubic_record_rtt() if the
mean is less than the min to ensure we have a sane mean for use in this
situation. This fixes the panic reported by Nick Hibma.
- Adjust conditions under which we update the mean RTT in cubic_post_recovery()
to ensure a low latency path won't yield an RTT of less than 1. This avoids
another potential divide by zero panic when running CUBIC in networks with
sub-millisecond latencies.
- Remove the "safety" assignment of min into mean when we don't update the mean
because of failed conditions. The above change to the conditions for updating
the mean ensures the safety issue is addressed and I feel it is better to keep
our previous mean estimate around if we can't update than to revert to the
min.
- Initialise the mean RTT to 1 on connection startup to act as a safety belt if
a situation we haven't considered and addressed with the above changes were to
crop up in the wild.
Sponsored by: FreeBSD Foundation
Reported and tested by: Nick Hibma
Discussed with: David Hayes <dahayes at swin edu au>
MFC after: 5 weeks
X-MFC with: r216114
value. While I'm here enable all clocks before initializing
controller. This change should fix lockup issue seen on AR8152
v1.1 PCIe Fast Ethernet controller.
PR: kern/154076
MFC after: 3 days
